LDL cholesterol: why it is the number that decides, and how low it is taken today

A lipid panel has five or six lines, and most people read all of them: total cholesterol, HDL, triglycerides, some ratio at the bottom. Yet in current guidelines the treatment decision is tied one way or another to a single line — low-density lipoprotein cholesterol, LDL-C. Not because the other numbers are useless: mechanism, genetics and randomised trials actually converge on all particles carrying apolipoprotein B. LDL ended up at the centre because it is what was deliberately lowered in the largest trials that counted outcomes (Ference et al., Eur Heart J 2017, PMID 28444290; ESC/EAS 2019, PMID 31504418).

Here is where that privileged position comes from, what levels LDL is lowered to today, what it is lowered with, and — most importantly — where the evidence stops.

Why LDL and not total cholesterol

Cholesterol does not float freely in blood; it travels inside lipoprotein particles. Every atherogenic particle — LDL, intermediate-density lipoprotein, VLDL remnants, lipoprotein(a) — carries exactly one molecule of apolipoprotein B (apoB) on its surface. Such a particle is small enough to cross the endothelium; once it is retained in the arterial wall, apoB binds to the matrix, the particle is oxidised, macrophages take it up — and that is how a plaque begins.

Two things follow. First, when the number of particles and the cholesterol inside them diverge, risk tracks closer to the particle count — which is why apoB, a direct particle count, is in many situations a sharper measure than LDL-C (PMID 31504418, 36216435). Second, total cholesterol mixes atherogenic particles together with HDL, which plays a different role in this story, and so it is a blunter basis for a decision.

The practical consequence behind the dry guideline text: when triglycerides are high, calculated LDL-C starts to mislead, and clinicians look instead at non-HDL cholesterol (total minus HDL) or at apoB. How these measures are built is covered in more detail in «Cholesterol and ApoB».

Three independent lines of evidence, not one

What makes LDL unusual is that three different kinds of data converge on it, and they do not contradict each other.

Observational cohorts. In large prospective studies, people with higher levels of atherogenic lipoproteins have more cardiovascular events. On its own this is an association, not an effect: people with high LDL differ in many other ways too.

Genetics. Mendelian randomisation gets around that limitation: gene variants are allocated at conception, essentially at random, so carriers of variants that confer lifelong lower LDL can be compared with everyone else as if in a natural randomised experiment. Such analyses — brought together in a consensus statement of the European Atherosclerosis Society (Ference et al., Eur Heart J 2017, PMID 28444290) — show an association proportional not only to the size of the LDL difference but also to its duration. That is the argument for cumulative exposure over years mattering more than the level at one moment.

Randomised trials with outcomes. Here one can speak of effect rather than association — and here it is essential to say in whom. The individual-participant analysis of 26 statin trials by the Cholesterol Treatment Trialists' Collaboration (Lancet 2010, PMID 21067804) covered roughly 170,000 people: trial participants, mostly middle-aged and older, most of them with either established cardiovascular disease or elevated calculated risk. In that population (169,138 participants), lowering LDL by 1.0 mmol/L was accompanied by a reduction in major vascular events of about 22% overall (RR 0.78; 95% CI 0.76–0.80): 12% in the first year, and roughly a quarter in each subsequent year. That is a relative reduction within groups of trial participants, not a forecast for any individual.

This matters for today's decisions: a similar proportionality per 1 mmol/L has been shown not only for statins but also for interventions that lower LDL by upregulating the LDL receptor — diet, bile acid sequestrants, partial ileal bypass, ezetimibe. CETP inhibitors, by contrast, did not deliver the expected reduction in events (Silverman et al., JAMA 2016, PMID 27673306). So the route to the number is not irrelevant: the proportionality holds where the lowering runs through the same mechanism.

🔴 What this picture does not say: it does not transfer automatically to an individual, and it does not answer what will happen to any one person. A proportional reduction in event rates within a group is a statement about the group; how much that amounts to for a given reader depends on their own absolute baseline risk — which is exactly why the decision belongs to a clinician and not to a table.

What level, and why documents disagree

There is no single "normal" number for everyone, and that is not an oversight — it is the design: the target depends on the risk category.

The 2019 ESC/EAS dyslipidaemia guidelines (Mach et al., Eur Heart J 2020, PMID 31504418) set a stepped framework: the higher the risk category, the lower the LDL target, down to below 1.4 mmol/L at very high risk, together with a requirement to lower the value by at least half from baseline. The 2018 American guideline (Grundy et al., Circulation 2019, PMID 30586774) is built differently: instead of hard targets it uses thresholds at which the next agent is added.

Hence a non-obvious but practically important consequence: two cardiologists working from different documents can name different numbers for the same lab result, and both are following the rules. The disagreement is not about the facts but about how risk is translated into a decision. If a specific target is named, it is reasonable to ask which risk category it came from and what went into that category.

What lowers it today

Below are the classes of intervention studied in trials with cardiovascular outcomes, and the populations they were studied in. This is not a recommendation to take anything: choice, combination and dose are a clinician's decision.

• Statins. The largest outcome evidence base; population — participants of 26 randomised trials, about 170,000 people, with established disease or elevated risk (CTT, PMID 21067804).
• Ezetimibe added to a statin. IMPROVE-IT: about 18,000 patients after an acute coronary syndrome; further LDL lowering and a modest further reduction in event rates (Cannon et al., NEJM 2015, PMID 26039521).
• PCSK9 inhibitors (evolocumab, alirocumab). FOURIER — 27,564 patients with established atherosclerotic disease on a statin (Sabatine et al., NEJM 2017, PMID 28304224); ODYSSEY OUTCOMES — 18,924 patients after an acute coronary syndrome (Schwartz et al., NEJM 2018, PMID 30403574).
• Bempedoic acid. CLEAR Outcomes: 13,970 people unable or unwilling to take a statin, a mix of primary and secondary prevention — a population for which outcome data did not previously exist (Nissen et al., NEJM 2023, PMID 36876740).
• Inclisiran (siRNA). In ORION-10 and ORION-11, LDL fell by roughly half (Ray et al., NEJM 2020, PMID 32187462); the cardiovascular outcome trials (ORION-4, VICTORION-2 Prevent) had published no results as of September 2026, and until they do, an effect on outcomes cannot be asserted.

Lipoprotein(a) is a separate line. Its level is largely genetically determined; statins do not lower it and, on trial data, may raise it slightly — which is clinically not a reason to stop a statin (EAS 2022, PMID 36036785). Drugs that lower Lp(a) specifically are in outcome trials, and as of September 2026 no event results have been published. So today Lp(a) is measured to sharpen the risk estimate, not to be treated in its own right.

Lower is better? Where the studied range ends

Recent trials have driven LDL down to values that until recently looked implausibly low, and within the populations and follow-up periods studied no additional harm was found — but that follow-up is a few years, not decades. What happens at a very low level over twenty or thirty years is simply unknown: nobody has watched for that long.

A separate thread is muscle complaints on statins. In blinded n-of-1 trials that alternated drug and placebo months within the same person — in participants who had previously stopped statins because of side effects — symptom intensity was on average comparable during placebo months: in SAMSON (n = 60) mean symptom intensity was 16.3 on the statin and 15.4 on placebo (Wood et al., NEJM 2020, PMID 33196154), and StatinWISE found no difference in muscle symptoms (Herrett et al., BMJ 2021, PMID 33627334). This does not mean the complaints are imagined: they are real, but in these groups their cause more often turned out not to be the drug itself. The conclusion is a clinician's to draw, and it is not "put up with it" but whether the link is worth re-testing.

What to do with the number

The practical frame looks like this. There is a measured number — LDL-C, better still alongside apoB and a once-in-a-lifetime Lp(a). There is a risk category, which is set not by the lipid profile alone but by the whole picture: blood pressure, smoking, diabetes, family history, vascular imaging where it exists. And there is a decision that follows from those two things — made by a clinician.

What a reader can do themselves is arrive at that conversation with numbers rather than impressions: see their own LDL series across several years rather than the single latest point, know whether Lp(a) has ever been measured at all, and understand which risk category they are placed in and why. A series is easiest to read on one chart — in Lonevi, each measure is assembled automatically from uploaded lab reports.

---

This material is informational and does not replace a consultation with a doctor. Decisions about therapy, targets and medicines are made by the treating clinician on the basis of the full clinical picture.